drag prediction using nsu3d (unstructured multigrid ns solver)
DESCRIPTION
Drag Prediction Using NSU3D (Unstructured Multigrid NS Solver). Dimitri J. Mavriplis National Institute of Aerospace Hampton, VA 23666 (USA). NSU3D Description. Unstructured Reynolds Averaged Navier-Stokes solver Vertex-based discertization Mixed elements (prisms in boundary layer) - PowerPoint PPT PresentationTRANSCRIPT
Drag Prediction Using NSU3D(Unstructured Multigrid NS Solver)
Dimitri J. Mavriplis
National Institute of Aerospace
Hampton, VA 23666 (USA)
NSU3D Description
• Unstructured Reynolds Averaged Navier-Stokes solver– Vertex-based discertization– Mixed elements (prisms in boundary layer)– Edge data structure– Matrix artificial dissipation
• Option for upwind scheme with gradient reconstruction
– No cross derivative viscous terms• Thin layer in all 3 directions
Solver Description (cont’d)
• Spalart-Allmaras turbulence model– (original published form)– Optional k-omega model
• Transition specified by zeroing out production term in turbulence model– More robust than S-A trip term– Based on surface patches
• Laminar patches• Laminar normal distance
Solution Strategy
• Jacobi/Line Preconditioning– Line solves in boundary layer regions
• Releives aspect ratio stiffness
• Agglomeration multigrid– Fast grid independent convergence rates
• Parallel implementation– MPI/OpenMP hybrid model
• DPW runs: MPI on 16 cpu cluster
Performance• Convergence in 500
multigrid cycles– Grid independent– Poor convergence for
negative alpha nacelle cases
• 16 Pentium IV 1.7GHz: Intermediate grids– 5 hours for 3M pt grid– 7.5 hours for 5M pt grid
Grid Generation
• Runs based on NASA Langley supplied VGRIDns unstructured grids
• Tetrahedra in Boundary Layer merged into prismatic elements
• Sequence of 3 grids for each configuration
• Transition based on surface patches
Grid Specifications
Configuration Coarse Medium Fine
Wing-Body
Points
Cells
1,121,301
6,558,758
3,010307
17,635,283
9,133,352
53,653,279
WB+Nacelle
Points
Cells
1,827,470
10,715,204
4,751,207
27,875,222
10,278,588
60,412,948
Intermediate Grid (WB: 3M pts)
• Transition specified by laminar patches
Intermediate Grid (WB: 3M pts)
• Transition specified by laminar patches
Intermediate Grid (WBN: 5M pts)
• Transition specified by laminar patches
Solution on Intermediate Grid (CL=0.5)
• Matching CL
Solution on Intermediate Grid (CL=0.5)
• Matching Incidence
Solution on Intermediate Grid (CL=0.5)
• Matching CL
Solution on Intermediate Grid (CL=0.5)
• Matching Incidence
Lift vs. Incidence (Wing-Body)
• Large Lift overprediction (incidence shift)
Surface Pressure at y/b=0.411(WB)
• Solution converges with grid refinement• Poor match with experiment for specified CL
• Better match for specified incidence
Lift vs. Incidence (WB-Nacelle)
• Large Lift overprediction (incidence shift)
Surface Pressure at y/b=0.411(WBNacelle)
• Solution converges with grid refinement• Poor match with experiment for specified CL
• Better match for specified incidence
Drag Polar (Wing-Body)
• Coarse grid inadequate• Correct shape, slight underprediction
Drag Polar (Wing-Body-Nacelle)
• Coarse grid inadequate• Correct shape, slight underprediction• Convergence issues at negative incidences
Moment (Wing-Body)
• Poor moment prediction– Related to incidence shift
Moment (Wing-BodyNacelle)
• Poor moment prediction– Related to incidence shift
Incremental Drag
• Within 2 to 4 counts at CL=0.5• Transition effects cancel out• More discrepancies at lower lift
Incremental Moment
• Good prediction in spite of poor absolute values– Cross over well predicted– Discrepancies at lower lift
Drag Rise Curves
• Medium grids, transition• General trend, Drag underpredicted• Cl vs incidence issues
Flow Details
• Separation bubble on WB-Nacelle (fine grid)
Flow Details
• Strong shock inboard pylon (-2 deg)• Ahead of specified transition region on lower wing
Conclusions
• CL – Incidence Issues
• Better success at prediction of increments
• Grid Generation approach may be more important than actual resolution
• Aero is still a tough problem– Worthy of funding